Coverage for /home/airflow/.local/lib/python3.12/site-packages/airflow/api_fastapi/core_api/services/ui/dag_run.py: 26%

1# Licensed to the Apache Software Foundation (ASF) under one 

2# or more contributor license agreements. See the NOTICE file 

3# distributed with this work for additional information 

4# regarding copyright ownership. The ASF licenses this file 

5# to you under the Apache License, Version 2.0 (the 

6# "License"); you may not use this file except in compliance 

7# with the License. You may obtain a copy of the License at 

8# 

9# http://www.apache.org/licenses/LICENSE-2.0 

10# 

11# Unless required by applicable law or agreed to in writing, 

12# software distributed under the License is distributed on an 

13# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY 

14# KIND, either express or implied. See the License for the 

15# specific language governing permissions and limitations 

16# under the License. 

17 

18from __future__ import annotations 

19 

20import statistics 

21from collections import Counter 

22 

23from airflow.api_fastapi.core_api.datamodels.ui.dag_runs import DurationStats 

24 

25 

26def compute_duration_stats(durations: list[float]) -> DurationStats | None: 

27 """ 

28 Compute duration statistics from a list of completed DAG run durations (in seconds). 

29 

30 Returns None when the list is empty (no completed runs exist yet). 

31 Mode is computed on second-rounded values to avoid float precision noise; returns None 

32 when every run has a unique duration. 

33 Percentiles use linear interpolation between adjacent sorted values. 

34 """ 

35 if not durations: 

36 return None 

37 

38 # On Postgres 14+, DagRun.duration comes back as Decimal (EXTRACT(epoch ...) returns 

39 # numeric); coerce to float so the percentile interpolation below (Decimal * float) works. 

40 sorted_d = sorted(float(d) for d in durations) 

41 

42 counts = Counter(round(d) for d in sorted_d) 

43 max_count = max(counts.values()) 

44 mode_val: float | None = ( 

45 float(min(k for k, v in counts.items() if v == max_count)) if max_count > 1 else None 

46 ) 

47 

48 def _percentile(p: float) -> float: 

49 idx = (len(sorted_d) - 1) * p / 100 

50 lo = int(idx) 

51 hi = min(lo + 1, len(sorted_d) - 1) 

52 return sorted_d[lo] + (sorted_d[hi] - sorted_d[lo]) * (idx - lo) 

53 

54 return DurationStats( 

55 mean=round(statistics.mean(sorted_d), 3), 

56 mode=round(mode_val, 3) if mode_val is not None else None, 

57 p50=round(_percentile(50), 3), 

58 p90=round(_percentile(90), 3), 

59 p95=round(_percentile(95), 3), 

60 p99=round(_percentile(99), 3), 

61 )